Perplexity earns its 8.7 on our 361 score for one reason: it shows its sources. That is a genuine structural advantage over a chat interface that asserts things. It is also the exact feature most likely to make you lazy.
What it is genuinely good at
- Finding the shape of a topic fast when you do not yet know the right vocabulary to search with.
- Surfacing primary sources you would not have found through a keyword search.
- Checking whether a claim you already have is actually supported anywhere.
The failure mode nobody admits to
A cited answer feels verified. It is not. The citation tells you a source exists and was consulted — it does not guarantee the source says what the summary claims it says, and it does not tell you the source is any good.
We hit this while researching this very week's news: one widely-syndicated article stated flatly that a model had been released, and a second, equally confident piece said it had not. Both had sources. Only the vendor's own newsroom settled it.
The discipline that makes it trustworthy
- Open the citation before you use the claim. Every time, for anything you will publish.
- Prefer the primary source over the aggregator — a company changelog beats an article about the changelog.
- Treat dates as claims that need checking too. Stale posts about fast-moving topics are the single most common source of confident error.
- When two sources disagree, do not average them. Go find the one that has standing to settle it.
Where it does not belong
Perplexity is a research tool, not a drafting tool. Its output is optimised to be defensible, which is not the same as being readable. Take the findings, then write the piece somewhere else.